All OpenAI models, always current

OpenAI's strongest reasoning model. The default when accuracy beats latency.

Latest flagship general-purpose model. Default for new production builds.

OpenAI's current image model. Better text rendering than DALL·E 3 and supports edits.

Cheap workhorse. Great default for chatbots and bulk classification.

Smallest GPT-5.4. For short classification, simple completion, embeddings-adjacent tasks.

Prior-gen pro model. Still excellent — slightly cheaper than 5.5-pro per token.

Mainline 5.4 — strong general model at moderate cost.

Low-latency voice + text streaming. For phone agents and live-audio interfaces.

Code-specialized model with fill-in-the-middle. The pick for IDE plugins and code agents.

Late-2025 pro tier. Still strong for deep-reasoning workloads.

Mid-tier 5.x. Cheaper than 5.5 with most of the capability for chat workloads.

The original GPT-5 launch. Most clients have moved up to 5.4+; pinned here for legacy users.

Cheap reasoning model. Use for chain-of-thought tasks where 5.5-pro is overkill.

Strong reasoning model from the o-series. Math, code analysis, structured planning.

Long-context 4.x. Useful when you specifically need 1M tokens without going to 5.5.

Highest-quality OpenAI embeddings. The default for new RAG pipelines.

Multilingual speech-to-text. The canonical transcription model.